Nicolas Ferre wrote:

For platforms that implement a hardware watchdog, call its initialization routine in init_sequence. This location has been chosen to be the closest to initialization of console as some watchdog drivers are writing status messages. On the other hand, watchdog setup should be close to chip startup to be able to begin the triggering quickly.

I am not opposed to putting watchdog into the init sequence.

A current ARM way for initializing the watchdog is to do it in s_init. Why doesn't this work for you ?

Reviewing other watchdog additions in the init sequences from powerpc,sh,m68k If an init is to be added, this setup should be cut-n-pasted from one of these. There should be follow-on patches to convert the ARM s_init's over to using the init sequence.

Tom

Signed-off-by: Nicolas Ferre nicolas.ferre@atmel.com

arch/arm/lib/board.c | 7 +++++++ 1 files changed, 7 insertions(+), 0 deletions(-)

diff --git a/arch/arm/lib/board.c b/arch/arm/lib/board.c index f5660a9..569009e 100644 --- a/arch/arm/lib/board.c +++ b/arch/arm/lib/board.c @@ -70,6 +70,10 @@ extern int AT91F_DataflashInit(void); extern void dataflash_print_info(void); #endif

+#ifdef CONFIG_HW_WATCHDOG +extern void hw_watchdog_init(void); +#endif

#ifndef CONFIG_IDENT_STRING #define CONFIG_IDENT_STRING "" #endif @@ -251,6 +255,9 @@ init_fnc_t *init_sequence[] = { serial_init, /* serial communications setup */ console_init_f, /* stage 1 init of console */ display_banner, /* say that we are here */ +#if defined(CONFIG_HW_WATCHDOG)

• hw_watchdog_init, /* watchdog setup */

+#endif #if defined(CONFIG_DISPLAY_CPUINFO) print_cpuinfo, /* display cpu info (and speed) */ #endif